Abstract
Similar samples from dry horse feed and dry cat food were differentiated by a simple protocol that relies on the acquisition of 1D 1H NMR spectra of aqueous extracts at increasing soaking times and a posterior multivariate analysis of the data. The sequential withdrawal of aliquots from the same samples increased the significance of the statistical models that detect differences among the samples. These differences were identified by the backscaled coefficients plots of supervised models, like orthogonal projection to latent structure discriminant analysis, and evaluated depending on the weight of the contribution to the class separation. Statistical total correlation spectroscopy was applied to the data and gave insight into the components of the aqueous extracts, especially in regards to the molecules responsible for the class separation.
Keywords: Food analysis, multivariate data analysis, NMR, O-PLS-DA, PCA, statistical total correlation spectroscopy, wet extraction.